It occurs to me that this model works best (or maybe only works at all) where the data store’s view of the world is relatively small. For example, there are lots and lots of health care providers or retailers out there, but I only deal with a fairly small number of them. That means I can reasonably keep track of who has access to the sensitive parts of my data store

Some parts, e.g., the blog-like parts, might be open to the public at large. That’s also easy to track: “Anybody and everybody” is basically one list item. Or “people in my friends list” is also one item. Things get dicey when I have to keep track of, say, dozens or more items on the access list. Defining groups definitely helps, as long as I don’t spend all my time managing the groups …